Somers's d statistics

Somers's d is a nonparametric statistical measure used to determine the strength and direction of association between two ordinal variables. Developed by Robert H. Somers in 1962, its key feature is being 'asymmetric,' meaning it distinguishes between an independent (predictor) variable and a dependent (outcome) variable. Within risk management, its application aligns with statistical techniques suggested in ISO 31010:2019 (Risk management — Risk assessment techniques). Unlike symmetric measures like Kendall's tau-b, Somers's d is ideal for answering predictive questions such as, 'How well does risk factor X predict risk impact Y?' This makes it a powerful tool for quantifying the predictive relationship between risk drivers and their consequences.

Somers's d is applied in ERM through a structured, data-driven process: 1. **Data Definition and Collection**: Identify and define ordinal variables, such as 'supplier dependency' (low, medium, high) as the independent variable and 'supply chain disruption impact' (minor, moderate, severe) as the dependent variable. Collect paired data through surveys or incident logs. 2. **Calculation and Interpretation**: Use statistical software to compute the Somers's d value, which ranges from -1 to +1. A high positive value indicates that an increase in the independent variable predicts an increase in the dependent variable. 3. **Decision-Making and Prioritization**: Based on the results, management can prioritize controls. If 'cybersecurity training frequency' shows a strong negative d-value with 'data breach incidents,' it provides quantitative justification to increase training investment. This approach helps companies move from subjective assessments to evidence-based risk mitigation, leading to measurable outcomes like a 20% reduction in specific risk events.

Taiwan enterprises often face three key challenges when implementing Somers's d: 1. **Lack of Structured Data**: Many firms, especially SMEs, have risk data in unstructured text formats, making statistical analysis impossible. The solution is to implement a standardized digital risk register based on ISO 31000, using consistent ordinal scales (e.g., 1-5) for impact and likelihood. 2. **Statistical Skills Gap**: Risk management teams may lack the expertise to correctly apply and interpret statistical models. This can be overcome through targeted training from expert consultants and by using business intelligence (BI) tools with built-in statistical functions. 3. **Cultural Reliance on Intuition**: A management culture that prioritizes experience over data can be a significant barrier. The best approach is to conduct a small-scale proof-of-concept project, demonstrating how the analysis validates or refines expert judgment on a well-known risk, thereby building trust in the methodology.

Winners Consulting specializes in Somers's d statistics for Taiwan enterprises, delivering compliant management systems within 90 days. Free consultation: https://winners.com.tw/contact